This AI Tests Skincare on Skin That's Still Alive

Key Takeaways

• Four weeks vs. ten days: Outer Bio says its Yuna platform keeps surgical-leftover human skin alive and measurable for four weeks. Genoskin, a commercial vendor of live surgical skin, advertises up to 10 days for its NativeSkin model.
• Every number is the company’s own: 300+ donors, 10,000 treatments, 10 terabytes, a 5.9-million-compound screening library. The science rests on one bioRxiv preprint, dated April 30, 2026, which has not been peer reviewed as of August 26, 2026.
• Lady Gaga sits on the board; her partner Michael Polansky is CEO. Her own brand, Haus Labs, shares a scientist with the lab and has run joint projects with it.
• The number that decides everything is 120 to 130: by Polansky’s count, that is roughly how many active ingredients the Food and Drug Administration (FDA) has approved across every over-the-counter skin drug. Where an AI-found molecule goes after the lab depends on a rule most of this week’s coverage skipped.

A Month of Living Skin, From Surgery Leftovers

On the evening of August 21, 2026, Lady Gaga posted to X: “Meet @outerbio and YUNA. We are building the future of skin health.” She signed off, “Welcome from Michael, Me and the entire team at Outer Bio.” About forty minutes earlier, at 6:31 p.m. Pacific, TechCrunch had published a long lunch interview with Michael Polansky, her partner, describing him as “running an outfit that has spent years figuring out how to keep living human tissue alive outside the body” and doing it “without anyone outside the company knowing.” Three days later, on August 24, a Business Wire release out of Cambridge, Massachusetts made it official: Outer Bio was out of stealth.

Strip the celebrity away and the claim is still strange. Outer Bio does not grow skin. It takes real human skin that “would otherwise be discarded after surgery — mostly plastic surgery,” sourced through biobanks operating under “institutional review board oversight and documented donor consent,” principally the National Disease Research Interchange and the Cooperative Human Tissue Network. Then it keeps that skin alive. Not cells in a dish: full-thickness tissue, “epidermis, dermis, and resident immune compartment,” held “for four weeks, with native architecture preserved.”

Four weeks is the whole pitch. In Polansky’s own launch essay, the industry’s standard window for living skin outside a body is “the field’s three-to-seven-day window.” Wing Venture Capital, which led Outer Bio’s funding, puts it more bluntly in its investment memo: “when taken out of the body, human skin dies in 3-7 days.” A week is long enough to learn whether a chemical kills skin. It is not long enough to watch collagen rebuild, pigment shift, or a barrier heal, processes TechCrunch summarizes as ones “that take weeks to unfold.”

The reason this is an Artificial Intelligence (AI) story and not just a biology story is what the company does with the extra three weeks. Each treated skin sample is measured repeatedly, and “each compound-sample pair generates more than 30,000 data points.” That data trains a model that predicts which untested chemicals might help skin. The model’s picks go back onto living tissue, and the results retrain the model. TechCrunch’s description of the business is the most honest one available: “That loop is the actual product Outer Biosciences is selling right now: not a single skincare ingredient, but a faster way of finding one.”

How You Keep Skin Alive Outside a Body

Skin dies outside a body for a boring reason. It has no blood supply. Nutrients stop arriving, waste stops leaving, and the tissue starves in its own byproducts. Outer Bio’s answer, per TechCrunch, is “a proprietary support system developed by his team” that “feeds the tissue nutrients and removes metabolic waste, extending its viable life well past the industry norm.” The company describes the hardware, “a proprietary, highly porous engineered 3D-printed support structure designed to support air-liquid interface culture and maximize mass transfer through full-thickness tissue,” but not the chemistry of what flows through it.

What has been published are the limits. Outer Bio’s own launch page states that Yuna “does not have full vascular perfusion, which bounds what can be modeled about flow-dependent biology,” and that “it does not recapitulate the recruitment of circulating immune cells.” In plain terms: no blood flows through the tissue, and no new immune cells can arrive from elsewhere in the body the way they would in a real inflammatory response. The immune cells that were already living in the skin on the day of surgery are the only ones in the experiment. The Next Web adds a further hedge from the company: “Outer says a month-old sample still resembles a day-one sample without being identical to it.”

Here is how the field’s living-skin options compare, using each vendor’s own numbers.

The numbers matter because the marketing sometimes rounds them. Wing’s “3-7 days” describes ordinary ex vivo skin, the field’s default. Genoskin, a commercial vendor of the same kind of live surgical skin, does not claim seven days; its product page says “up to 10 days of testing on real human skin.” So the honest comparison is 28 days against 10, a 2.8-times advantage, not 28 against 7. That is still large. It is just not the number in the pitch decks.

What the Skin Was Made to Do

A biology claim is only as good as the experiments behind it, and Outer Bio has put three on the record. They appear in a preprint its scientists submitted to bioRxiv on April 30, 2026 (posted May 5), titled “Long-Term Human Skin Platform for Modeling Chronic Inflammation, Environmental Stress, and Therapeutic Intervention,” whose abstract describes a platform that “preserves native tissue architecture and epidermal, stromal, and immune-associated molecular programs for up to 4 weeks.” A preprint is a paper posted before peer review; as of this article’s date no reviewed version exists, and the author list closes with the company’s chief scientist and its chief technology officer in the senior-author slots. Treat every result below as Outer Bio’s own account.

Psoriasis in a dish. The company says “sustained IL-17/IL-22 stimulation produced a psoriasis-like inflammatory state across multiple donors — transcriptional, histological, and secretory — across three weeks, and tofacitinib, a JAK inhibitor used clinically, broadly suppressed key inflammatory and tissue-damage-associated features across the time course.” IL-17 and IL-22 are interleukins, the signaling proteins that drive psoriasis in real patients. Tofacitinib is a Janus kinase (JAK) inhibitor, a real prescription drug. If the result holds, it means the platform can model a chronic disease and then reverse it with a known medicine, which is a drug-development experiment, not a cosmetics one.

Aging skin. “On donor tissue from individuals aged 58-74, prolonged treatment with dasatinib and quercetin attenuated senescence and inflammaging signatures.” Dasatinib is a leukemia drug; quercetin is a plant flavonoid. Together they are the best-known “senolytic” pair, compounds meant to clear worn-out cells. Again: a pharmacology experiment.

Sunburn. “A controlled UVB exposure produced coordinated injury and inflammatory signaling that topical sunscreen substantially mitigated.” Ultraviolet B (UVB) is the wavelength that burns. Note the adverb: the launch page says “substantially,” while the preprint’s own abstract says the UVB response was “partially mitigated using topical sunscreen.” Same experiment, softer word in the document scientists will actually read. Of the wider list of stressors the company says the platform supports (pollution, microbes, mechanical force, wounding), UV is the only one it has written up so far: “The UV response is the one we’ve published; the others are supported in the platform, and we’ll publish more on them in the coming months.”

The scale behind those experiments, again in the company’s own words: “pooled data from more than 300 donors across the program to date,” spanning “Fitzpatrick skin types I through VI,” with “a single surgical specimen” yielding “up to a hundred biopsies.” The press release adds that “the platform has produced over 10 terabytes of proprietary data across more than 300 donors and 10,000 treatments,” and that its “virtual screening library already spans 5.9 million compounds across 11 databases.” Ten terabytes (TB) across 10,000 treatments works out to roughly one gigabyte per treatment, which is consistent with repeated molecular profiling rather than a few photos and a score.

Output, per Polansky: “the company is now generating a new candidate roughly every six weeks, with six leads currently active in its pipeline and several dozen additional ‘hits’ logged,” and “four of Outer Biosciences’ six current leads look likely to reach commercialization.”

The Part Nobody Mentions: Nothing Here Needs FDA Approval

Every headline this week said some version of “AI discovers skincare.” Almost none said what happens to a discovery once the AI makes it. TechCrunch did, in one sentence: “Outer Biosciences is discovering cosmetic ingredients, not drugs, so there’s no FDA approval to seek out.”

That is not a loophole Outer Bio found. It is the law, and the FDA states it plainly on its own site: “Under the law, cosmetic products and ingredients do not need FDA premarket approval, with the exception of color additives.” The 2022 Modernization of Cosmetics Regulation Act (MoCRA) added facility registration, product listing, and a deadline for Good Manufacturing Practice (GMP) rules, but the FDA’s explainer on cosmetics versus drugs is explicit that the approval gap survived: drugs “must generally either receive premarket approval by FDA through the New Drug Application (NDA) process or conform to a ‘monograph,’” while cosmetics need neither.

What decides which side of the line a product lands on is not the ingredient. It is the sentence on the box. “Whether a product is a cosmetic or a drug under the law is determined by a product’s intended use,” and “certain claims may cause a product to be considered a drug, even if the product is marketed as if it were a cosmetic.” Say a cream “beautifies” and it is a cosmetic. Say it “treats” or affects “the structure or any function of the body” and it is a drug, with the NDA and the clinical trials that follow. As for the industry’s favorite word for products that live in between, the FDA has a one-line ruling: “the term ‘cosmeceutical’ has no meaning under the law.”

Now read the psoriasis and senolytic experiments again. Those are disease and structure-of-the-body results. The moment an Outer Bio ingredient is sold with that language on the label, it becomes a drug and the FDA gate slams shut. Sold as a cosmetic, with cosmetic claims, the same molecule needs an ingredient name and a safety file and nothing else. This is why Polansky’s own framing of the opportunity is worth quoting in full. TechCrunch reports that across the FDA’s 13 categories of over-the-counter skin drugs, “only about 120 to 130 active ingredients are approved,” and that adding “cosmetic ingredients backed by actual research, says Polansky,” brings “that number is closer to 200.” In his launch essay he goes further: “In thirty years, topical skincare has produced exactly one FDA-approved mechanism for visible skin aging: retinoids.”

The 200 is his figure; the 120 to 130 is TechCrunch’s own count, and neither is an independent audit. But if they are close, the strategy is clear. A tool that can run drug-grade experiments has been aimed, to begin with, at the one product category where a positive result does not need to survive a regulator. That is the quickest path to revenue, and there is nothing improper about it. It just means “AI-tested” on a future jar will describe how the ingredient was found, not whether anyone was required to prove it works on your face.

Who Gets Paid

Follow the money in every direction, and the picture is ordinary rather than sinister. Outer Bio “has raised roughly $23 million, with early backers including Wing Ventures, Initialized, and Polansky’s own investing firm, Hawktail, among others,” and “employs 19 people, with all but Polansky based just outside Cambridge, Massachusetts.” The press release lists the full syndicate: “Wing Venture Capital, Initialized Capital, Sozo Ventures, Hawktail, Lightspeed Venture Partners, Third Kind Venture Capital, Liquid 2 Ventures, and SV Angel.” Polansky is the CEO, an investor through Hawktail, and, in TechCrunch’s words, “the creative, business, and romantic partner to Stefani Germanotta,” the singer’s legal name. Germanotta “also sits on the board of Outer Biosciences,” owns Haus Labs, a cosmetics brand of “roughly 70 employees,” and “Outer Biosciences’ chief scientist, Kyung-Jin Jang, sits on Haus Labs’ scientific advisory board, and the companies have run some joint projects.” For the record on titles: TechCrunch names the four co-founders as Polansky, Jang, Chris Hinojosa and Stanley King, Outer Bio’s own launch essay is bylined “Michael Polansky, CEO & Founder,” and Germanotta is a board member, not a founder.

The business model is licensing: Outer Bio “right now plans to license or sell its finished ingredients to beauty or pharma outfits that will then formulate these into actual products (a serum, a cream) and bring them to market under their own brands.” A brand whose founder sits on the board and whose advisory board includes the lab’s chief scientist is an obvious candidate to become a customer, though no such deal has been announced and TechCrunch describes the two companies as collaborating “at the margins.” Revenue for now comes from “collaborative research partnerships, including a pharmaceutical partner that’s studying why certain cancer drugs cause severe skin rashes, and consumer beauty brands that are testing whether Outer Biosciences’ data holds up.” Note the word “pharmaceutical.” The drug-grade experiments and the drug-company partner are the tell that cosmetics are the beachhead, not the destination.

What Would Change the Story

The case against taking any of this at face value is straightforward, and the company has partly made it for you. The only paper is a preprint that has not passed peer review, whose corresponding author is at Outer Biosciences. The Next Web, one of the few outlets to note the hedging, wrote of the company’s day-one-resemblance claim: “That is a carefully hedged claim and the company makes it carefully.” No outside skin biologist has been quoted validating the platform in any coverage this article could find. The 300 donors, 10,000 treatments and 5.9 million compounds are all self-reported.

The competitive field is also older than the headlines suggest. The European Union (EU) banned marketing cosmetics tested on animals in stages ending on March 11, 2013, “irrespective of the availability of alternative non-animal tests.” L’Oréal’s reconstructed-skin program predates that ban by decades: its own timeline dates its “1st reconstructed human epidermis model” to 1979 and its decision to stop animal testing to 1989, and its EpiSkin model carries Organisation for Economic Co-operation and Development (OECD) validation for phototoxicity (2002), with EpiSkin and SkinEthic both validated for skin irritation (2010). Genoskin sells immunocompetent live skin as of August 2026. And TechCrunch notes a better-funded rival in the broader “living human tissue plus AI” space: Vivodyne, “a Philadelphia-based competitor building lab-grown human organ tissue,” which the outlet reports has “raised close to $80 million to date.” Outer Bio is not the first company to keep human skin alive for research. Its claim is to keep it alive longer, and to close the loop with a model.

So what would move this from “interesting claim” to “real”? Three things, in order of weight. A peer-reviewed version of the preprint, with day-zero versus day-28 data that outside labs can inspect. An independent replication, ideally by a customer with no equity. And a licensed ingredient reaching a shelf with a label that says what the lab actually showed rather than what the cosmetics rulebook allows it to say.

The biology claim and the incentive stack are not equally settled. The incentive stack is fully public and fully ordinary. The four-week skin is a company claim still waiting for an outside witness. The reader who sees “AI-tested” on a serum in 2027 should ask which of the two they are buying.

The next checkable date is whichever comes sooner: a peer-reviewed journal accepting the April preprint, or a product page anywhere that names an Outer Bio ingredient. For a sense of what happens when a skin-adjacent molecule does go the full drug route, with trials and an FDA decision at the end, compare the path Eli Lilly’s retatrutide took to a Phase 3 readout. Same word, “works,” two very different burdens of proof.

Sources (14)

• outerbio.com Introducing Yuna
• outerbio.com Why we are building Outer Bio
• finance.yahoo.com Business Wire: Outer Bio Emerges from Stealth with Yuna
• biorxiv.org bioRxiv preprint: Long-Term Human Skin Platform for Modeling Chronic Inflammation, Environmental Stress, and Therapeutic Intervention
• wing.vc Reengineering Skin Health with Yuna
• techcrunch.com Michael Polansky is training an AI model on skin that's still alive
• thenextweb.com This startup is training an AI model on skin that stays alive for a month
• x.com Lady Gaga on X: Meet @outerbio and YUNA
• genoskin.com NativeSkin human skin model
• fda.gov FDA Authority Over Cosmetics
• fda.gov Is It a Cosmetic, a Drug, or Both?
• fda.gov Cosmeceutical
• single-market-economy.ec.europa.eu European Commission: Ban on animal testing
• loreal.com L'Oréal: Milestones in non-animal safety testing